A benchmark for evaluating whether large language models can generate complete, logically rigorous proofs for Olympiad-level mathematics problems.
Browse the latest scores, model modes, release dates, and parameter sizes for IMO-ProofBench.
Data sourced primarily from official releases (GitHub, Hugging Face, papers), then benchmark leaderboards, then third-party evaluators. Learn about our data methodology
| Rank | Model | License | |||
|---|---|---|---|---|---|
 DeepSeekMath-V2 Thinking Level · High | 99.00 | 2025-11-27 | 685B | Free Commercial | |
 GPT-5 Thinking Enabled | 59.00 | 2025-08-07 | Unknown | Closed | |
 Gemini 2.5-Pro Thinking Enabled | 55.20 | 2025-06-05 | Unknown | Closed | |
4 |  Grok 4 Thinking Enabled | 46.70 | 2025-07-10 | Unknown | Closed |
5 | Gemini 2.5 Deep Think Thinking Enabled | 37.60 | 2025-08-01 | Unknown | Closed |
6 |  Qwen3-235B-A22B-Thinking Thinking Enabled | 33.30 | 2025-07-30 | 30.5B | Free Commercial |
7 | DeepSeek-R1-0528 Thinking Enabled | 29.00 | 2025-05-28 | 671B | Free Commercial |
8 |  Claude Sonnet 4 Thinking Enabled | 27.10 | 2025-05-23 | Unknown | Closed |
9 | Claude Sonnet 4.5 Thinking Enabled | 27.10 | 2025-09-30 | Unknown | Closed |
10 | Grok 4 Deep Thinking Mode | 23.30 | 2025-07-10 | Unknown | Closed |
11 | OpenAI o3 Thinking Enabled | 20.50 | 2025-04-16 | Unknown | Closed |
12 | OpenAI o4 - mini Thinking Level · High | 11.40 | 2025-04-16 | Unknown | Closed |
13 |  Kimi K2 0905 Standard Mode | 7.10 | 2025-09-05 | 1000B | Free Commercial |
14 | Qwen3-VL-235B-A22B-Instruct Standard Mode | 5.20 | 2025-09-23 | 235B | Free Commercial |
15 | DeepSeek-V3-0324 Standard Mode | 4.30 | 2025-03-24 | 671B | Free Commercial |
16 | Claude Opus 4 Thinking Enabled | 2.90 | 2025-05-23 | Unknown | Closed |
